TY - JOUR T1 - Linear versus nonlinear regression for quantification of dynamic PET with amphetamine challenge JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 160P LP - 160P VL - 48 IS - supplement 2 AU - Yun Zhou AU - Michael Weed AU - Arman Rahmim AU - Weiguo Ye AU - James Braši\#263; AU - Mohab Alexander AU - Andrew Crabb AU - Farah Ali AU - Dean Wong Y1 - 2007/05/01 UR - http://jnm.snmjournals.org/content/48/supplement_2/160P.2.abstract N2 - 546 Objectives: To evaluate a linear regression algorithm for the quantification of dynamic PET with amphetamine challenge using an improved ESRTM model. Methods: Sixteen Rhesus monkeys were studied with [11C]raclopride delivered by bolus plus continuous infusion of Kbol = 75 min. PET scanning started at the beginning of tracer injection on a high resolution research tomography (HRRT) scanner. Forty min post tracer injection, amphetamine (2 mg/kg) was injected intravenously over 2 min. The dynamic images of 30 and 34 frames were reconstructed for 2 typical PET temporal sampling. A parameter dt that represents a latent period in the amphetamine-induced displacement of tracer with the ESRTM of parameters (R1, k2, BP0, BP1) (Zhou et al., Neuroimage 2006, 33(2):550-63) (ESRTMdt) was proposed for modeling tracer kinetics in the baseline phase [0 40+dt] and the displacement phase [40+dt 90]. The following linear regression algorithm to fit ESRTMdt was implemented: 1) Sample dt in [0 30] with step size 0.5 min. 2) Estimate R1, k2, BP0, BP1 by linear regression and calculate residual sum squares (RSS) with sampled dt values and a linear operational equation for ESRTMdt. 3) Utilize the R1, k2, BP0, BP1 and dt values corresponding to the least RSS as the final linear estimates. A nonlinear regression method using the Marquardt algorithm was used to estimate 5 parameters of ESRTMdt simultaneously. Results: The estimates of (R1, k2, BP0, BP1, dt) obtained by fitting ESRTMdt to the measured striatum time activity curves of 30 time points were (0.95 ± 0.09, 0.18 ± 0.02, 4.79 ± 0.69, 3.43 ± 0.59, 14.97 ± 6.69) and (0.95 ± 0.09, 0.18 ± 0.02, 4.81 ± 0.68, 3.42 ± 0.61, 14.78 ± 6.47) (mean ± SD, n = 16) for linear and nonlinear regression, respectively. The estimates from the two typical PET temporal sampling data were similar for both linear and nonlinear regression methods. Conclusions: The proposed linear regression method is as reliable as the nonlinear regression approach for region of interest based kinetic analysis with the ESRTMdt model. Its application for linear parametric imaging with ESRTMdt is under investigation. Research Support (if any): Grant support AA12839, DA00412, NS38927, MH075378 ER -